How to Interpret Variants in Non-Coding Regions: Applying ACMG/AMP Criteria to Regulatory and Splice-Site Variants

By Dr. Zubair Khalid, DVM, MS, PhD · 2026-08-21

Key Takeaways

• Non-coding variants necessitate a distinct interpretation strategy from coding variants due to their indirect functional impact on gene regulation and splicing, requiring integration of population frequency, in silico predictions, and functional validation within the ACMG/AMP framework.
• Canonical splice site variants (±1, ±2) have the most direct evidence pathway, often classified as pathogenic or likely pathogenic when functional data (e.g., RT-PCR analysis of aberrant splicing) confirm aberrant splicing, while deep intronic variants typically remain VUS without such validation.
• Functional evidence, particularly from assays like minigene splicing assays or reporter gene assays, is paramount for non-coding variant classification, providing direct mechanistic insight that in silico tools alone cannot achieve.
• Comprehensive annotation including gene context, genomic location, conservation, regulatory element overlap, and population frequency (e.g., gnomAD) is the foundational step, followed by in silico prioritization and targeted functional validation.
• Whole genome sequencing offers broader detection of non-coding variants but increases the interpretation burden; RNA sequencing provides physiological relevance for splicing impact, while minigene assays offer experimental control.
• Common failure patterns include overreliance on in silico predictions, neglecting tissue-specific effects, incomplete population frequency assessment, and ignoring cryptic splice site activation, leading to misclassification.

Non-coding variants present a distinct interpretation problem for laboratory professionals and researchers. Unlike missense or frameshift variants in coding exons, variants in promoters, enhancers, 5' and 3' untranslated regions, deep intronic sequences, and splice consensus sites do not directly alter protein sequence. Their functional impact depends on regulatory context, tissue-specific expression patterns, and splicing machinery recognition. The American College of Medical Genetics and Genomics (ACMG) and Association for Molecular Pathology (AMP) criteria were designed primarily with coding variants in mind, yet they remain the standard framework for classifying all variant types. This article provides a practical framework for applying ACMG/AMP criteria to non-coding variants, with emphasis on functional validation, population frequency filtering, and disease-specific database integration. The target reader is a bioinformatics practitioner, clinical laboratory scientist, or genetics researcher who needs concrete decision criteria for variant classification workflows.

Scope and Reader Context

The interpretation of non-coding variants requires a different evidentiary standard than coding variants because the functional consequence is often indirect. A splice donor variant may lead to exon skipping, intron retention, or cryptic splice site activation. A promoter variant may reduce transcription initiation efficiency without abolishing it entirely. A deep intronic variant may create a cryptic branchpoint that competes with the native splicing machinery. Each mechanism produces a different molecular phenotype, and each requires different experimental validation.

This article addresses the specific problem of classifying variants outside coding regions using the ACMG/AMP framework. The practical outcome is a workflow that integrates population frequency data, in silico prediction tools, functional assay evidence, and disease-specific knowledge to assign pathogenicity classifications. The workflow applies to germline variant interpretation in diagnostic laboratories, research settings, and carrier screening programs. Somatic variant interpretation follows different rules and is addressed only where relevant to germline classification.

The evidence base for this article includes peer-reviewed studies of non-coding variant interpretation in inherited retinal disease, carrier screening programs, and early-onset colorectal cancer. These studies demonstrate both the challenges and the successful strategies for classifying regulatory and splice-site variants. The National Center for Biotechnology Information (NCBI) provides the underlying sequence and variation databases, while the European Bioinformatics Institute (EMBL-EBI) offers training resources for the computational skills required. Bioconductor, Galaxy, and nf-core provide the reproducible analysis infrastructure.

At a Glance: Non-Coding Variant Classification Decision Table

The decision table above summarizes the typical classification pathways for four non-coding variant classes. Canonical splice site variants have the strongest evidence pathway because the splicing machinery has well-defined consensus sequences. Deep intronic variants require functional validation because in silico predictions alone are insufficient. Promoter and enhancer variants require tissue-specific functional data. UTR variants are the most challenging because their regulatory mechanisms are diverse and often poorly characterized.

The Non-Coding Variant Interpretation Problem

Why Non-Coding Variants Are Difficult to Classify

The ACMG/AMP guidelines were developed primarily for coding variants where the functional consequence can be predicted from the genetic code. A nonsense variant introduces a premature stop codon. A frameshift variant alters the reading frame. A missense variant changes an amino acid. These consequences are predictable from sequence alone, even if the phenotypic effect requires additional evidence.

Non-coding variants lack this direct predictability. A variant in a promoter may affect transcription factor binding, but the effect depends on which transcription factors bind that specific promoter, in which tissues, and under which conditions. A variant in a deep intronic region may create a cryptic splice site, but the effect depends on the local splicing regulatory environment, the strength of competing splice sites, and the expression level of splicing factors. A variant in a 3' UTR may disrupt a miRNA binding site, but the effect depends on which miRNAs are expressed in the relevant tissue and whether the target gene is regulated by that particular miRNA.

The clinical genomics community has recognized this challenge. In a study of inherited retinal diseases, researchers noted that the interpretation of variants of uncertain significance remains a barrier to molecular diagnosis, and they emphasized the role of functional assays in addressing unresolved cases. The study traced the evolution of testing from linkage analysis to high-throughput sequencing and highlighted that standardized variant interpretation frameworks have increased diagnostic yield, but non-coding variant interpretation remains an active area of method development [<a href="#ref-1">1</a>].

The Scale of the Problem in Clinical Sequencing

The volume of non-coding variants identified in clinical sequencing is substantial. A study of early-onset colorectal cancer in 125 patients from Kazakhstan identified 11,152 variants across 85 cancer-associated genes. Of these, 1,042 were intronic or non-coding variants, 44 were 3' UTR variants, two were splice donor variants, and one was a splice acceptor variant. The study applied ACMG guidelines and the LOVD and ClinVar databases to classify variants, identifying 24 pathogenic variants and 289 variants of uncertain significance with population frequency below 1% [<a href="#ref-2">2</a>].

This distribution illustrates the interpretation burden. The majority of non-coding variants identified in clinical sequencing are of uncertain significance. The study found that 50% of all pathogenic mutations in Kazakhstani patients with early-onset colorectal cancer were identified in subgroups with a family history of colorectal cancer and primary multiple tumors, demonstrating the importance of clinical context in variant interpretation [<a href="#ref-2">2</a>].

The Role of Genome Sequencing in Expanding Non-Coding Variant Detection

Genome sequencing has expanded the scope of non-coding variant detection beyond what targeted gene panels and exome sequencing can achieve. A study of preconception carrier screening using genome sequencing analyzed 728 gene-disorder pairs in 131 women and their partners. The study reported five variants in non-coding regions and ten copy-number variants. The authors noted that genome sequencing improves the sensitivity of detecting clinically significant variants compared with targeted mutation screening, while acknowledging that certain novel variant interpretation remains challenging [<a href="#ref-3">3</a>].

The expansion of sequencing to include non-coding regions creates both opportunities and challenges. The opportunity is the detection of pathogenic variants that would be missed by exome or panel sequencing. The challenge is the interpretation burden of the many non-coding variants that are identified but cannot be classified without additional evidence.

Core Principles of ACMG/AMP Criteria Applied to Non-Coding Variants

The Evidence Framework

The ACMG/AMP framework assigns pathogenicity based on evidence categories. Each category has a code, and combinations of codes determine the final classification. The framework is designed to be transparent and reproducible, with each evidence code requiring specific types of data.

For non-coding variants, the applicable evidence codes include:

• PVS1 (null variant in a gene where loss of function is a known mechanism of disease): For non-coding variants, PVS1 requires evidence that the variant causes loss of function through aberrant splicing, promoter disruption, or other mechanisms.
• PS1 (same amino acid change as an established pathogenic variant): This code applies to coding variants and has limited applicability to non-coding variants, though it may apply to splice-site variants that produce the same aberrant transcript as a known pathogenic variant.
• PS3 (well-established functional studies showing a deleterious effect): This code is critical for non-coding variants because functional assays provide the most direct evidence of impact.
• PM2 (absent from controls in population databases): This code applies to all variant types and is commonly used for non-coding variants.
• PP3 (computational evidence supporting a deleterious effect): This code applies to in silico predictions, including splicing prediction tools.
• BS3 (well-established functional studies showing no deleterious effect): This code is the benign counterpart of PS3 and requires functional evidence of normal function.

The application of these codes to non-coding variants requires careful consideration of what constitutes appropriate evidence for each category.

The Central Role of Functional Evidence

Functional evidence is the cornerstone of non-coding variant interpretation. The ACMG/AMP framework assigns the strongest weight to well-established functional studies, and for non-coding variants, functional studies are often the only way to move beyond a variant of uncertain significance classification.

A study of a cryptic branchpoint variant in the RPGR gene demonstrated this principle. The variant, RPGR NM_001034853.2 c.1307G>A, was identified in a large Irish pedigree with X-linked retinitis pigmentosa. In silico investigations using SpliceAI and Alamut Visual software suggested a potential splicing impact, but the classification remained uncertain until functional analysis was performed. The researchers used in vitro midigene splice assays with gateway expression vectors to interrogate the effect of the variant on RNA splicing. The assays confirmed that the variant created a cryptic acceptor site and a cryptic branchpoint motif, leading to the excision of intron 10 and 90 bases of exon 11. This aberrant splicing created a frameshift and a premature stop codon, with no functional RPGR transcript predicted to remain [<a href="#ref-4">4</a>].

The functional evidence allowed the researchers to upgrade the variant classification to pathogenic using ACMG/AMP and ClinGen Sequence Variant Interpretation recommendations. This case illustrates the critical role of functional assays in non-coding variant interpretation and demonstrates that even variants in exonic regions can have splicing effects that require functional validation [<a href="#ref-4">4</a>].

The Limitations of In Silico Prediction

In silico prediction tools are useful for prioritizing variants for functional analysis, but they are not sufficient for classification. The RPGR study used SpliceAI and Alamut Visual software to identify the variant as a candidate for functional analysis, but the classification was based on the functional assay results, not the predictions alone [<a href="#ref-4">4</a>].

The limitations of in silico prediction are particularly acute for non-coding variants because the prediction tools are trained on known splice sites and regulatory elements. Variants that create cryptic splice sites or branchpoints may not be well predicted by tools that focus on canonical splice site recognition. Variants in promoters and enhancers are even more challenging because the regulatory grammar is less well understood than the splice code.

The practical implication is that in silico prediction should be used for variant prioritization and hypothesis generation, not for final classification. A variant with strong in silico evidence of splicing impact should be referred for functional analysis if clinical interpretation is needed.

Practical Workflow for Non-Coding Variant Classification

Step 1: Variant Annotation and Contextualization

The first step in non-coding variant classification is comprehensive annotation. This includes:

• Gene and transcript context: Which gene and transcript does the variant affect? Is the variant in a canonical transcript or an alternative transcript?
• Genomic location: Is the variant in a promoter, enhancer, 5' UTR, 3' UTR, intron, or splice site?
• Conservation: Is the variant in a conserved region across species?
• Regulatory annotation: Does the variant overlap known regulatory elements such as transcription factor binding sites, DNase hypersensitivity sites, or histone modification marks?
• Population frequency: What is the allele frequency in population databases such as gnomAD?

The NCBI provides access to the underlying sequence and variation databases needed for this annotation. The NCBI resources include the Reference Sequence database for transcript information, dbSNP for variant information, and the Variation Services for querying variant data. These resources are essential for establishing the genomic context of a non-coding variant [<a href="#ref-5">5</a>].

The EMBL-EBI training resources provide instruction on using these databases effectively. The training materials cover data retrieval, sequence analysis, and variant annotation, providing the foundational skills needed for non-coding variant interpretation [<a href="#ref-6">6</a>].

Step 2: Population Frequency Filtering

Population frequency is a critical filter for non-coding variant classification. The PM2 criterion requires that a variant be absent from controls in population databases, and the BS1 criterion requires that a variant have a frequency that is too high for the disorder.

For non-coding variants, the same population frequency thresholds apply as for coding variants. A variant that is common in the general population is unlikely to be pathogenic for a rare Mendelian disorder. However, the interpretation of population frequency requires consideration of the disorder prevalence, the mode of inheritance, and the penetrance of the variant.

The carrier screening study provides an example of population frequency filtering in practice. The study filtered variants and classified them using the latest ACMG guidelines, reporting only pathogenic and likely pathogenic variants after confirmation by orthologous methods. Novel missense variants were classified as variants of uncertain significance, and novel splice-site variants underwent RNA-splicing assays to aid in classification [<a href="#ref-3">3</a>].

Step 3: In Silico Prediction and Prioritization

In silico prediction tools should be applied to non-coding variants to prioritize them for functional analysis. The tools include:

• Splice site prediction tools: SpliceAI, Alamut Visual, and similar tools predict the impact of variants on splicing.
• Regulatory element prediction tools: Tools that predict the impact of variants on transcription factor binding, chromatin state, and enhancer activity.
• Conservation-based tools: Tools that use evolutionary conservation to predict functional impact.

The RPGR study used SpliceAI version 1.3.1 and Alamut Visual software version 2.13 for in silico investigation. These tools identified the variant as a candidate for functional analysis, and the functional assays confirmed the predicted splicing impact [<a href="#ref-4">4</a>].

The practical approach is to use in silico prediction as a triage tool. Variants with strong in silico evidence of splicing impact should be prioritized for functional analysis. Variants with weak or no in silico evidence may still warrant functional analysis if they segregate with disease in a pedigree or if they are in a gene with a strong prior probability of pathogenicity.

Step 4: Functional Assay Design and Execution

Functional assays provide the most direct evidence of non-coding variant impact. The choice of assay depends on the variant location and the suspected mechanism:

• Splicing assays: For variants in or near splice sites, deep intronic variants, and exonic variants that may affect splicing. The RPGR study used midigene splice assays with gateway expression vectors. The variant and wildtype RNA were amplified by RT-PCR to investigate effects on splicing [<a href="#ref-4">4</a>].
• Reporter assays: For promoter and enhancer variants. These assays measure the effect of a variant on gene expression by cloning the variant sequence upstream of a reporter gene.
• RNA sequencing: For variants in genes expressed in accessible tissues. RNA sequencing can detect aberrant splicing, allele-specific expression, and nonsense-mediated decay.
• Protein expression assays: For variants that may affect translation efficiency or protein stability.

The carrier screening study used RNA-splicing assays for novel splice-site variants to aid in classification. This approach allowed the laboratory to classify variants that would otherwise remain of uncertain significance [<a href="#ref-3">3</a>].

Step 5: Evidence Integration and Classification

The final step is integrating all evidence into an ACMG/AMP classification. The evidence codes are applied based on the strength and quality of the evidence, and the combination of codes determines the classification.

For the RPGR variant, the evidence included:

• PVS1: The functional assay demonstrated that the variant leads to a frameshift and premature stop codon, with no functional transcript predicted to remain.
• PM2: The variant was absent from population databases.
• PP3: In silico prediction tools supported a splicing impact.
• PS3: The functional assay provided well-established evidence of a deleterious effect.

The combination of these evidence codes supported a pathogenic classification [<a href="#ref-4">4</a>].

Options and Tradeoffs in Non-Coding Variant Analysis

Whole Genome Sequencing Versus Targeted Approaches

Whole genome sequencing provides the most comprehensive view of non-coding variants, but it also generates the largest interpretation burden. The carrier screening study used genome sequencing to analyze 728 gene-disorder pairs and identified five variants in non-coding regions. The study noted that genome sequencing improves the sensitivity of detecting clinically significant variants compared with targeted mutation screening [<a href="#ref-3">3</a>].

Targeted approaches, such as gene panels and exome sequencing, have a lower interpretation burden but may miss non-coding variants. The choice between approaches depends on the clinical context, the gene list, and the likelihood of non-coding variant involvement.

RNA Sequencing Versus Minigene Assays

RNA sequencing provides the most physiologically relevant assessment of splicing impact because it measures the actual transcript in the relevant tissue. However, RNA sequencing requires access to the relevant tissue, and many genes are not expressed in accessible tissues such as blood.

Minigene assays provide a controlled experimental system for assessing splicing impact. The RPGR study used midigene splice assays with gateway expression vectors, which allowed the researchers to assess the splicing impact of the variant in a controlled system [<a href="#ref-4">4</a>]. Minigene assays are particularly useful when the relevant tissue is not accessible for RNA sequencing.

The tradeoff is between physiological relevance and experimental control. RNA sequencing provides physiological relevance but may be limited by tissue access. Minigene assays provide experimental control but may not fully recapitulate the native splicing environment.

In Silico Prediction Tools

The choice of in silico prediction tools involves tradeoffs between sensitivity and specificity. Tools that are highly sensitive may generate many false positives, while tools that are highly specific may miss true positives.

The RPGR study used multiple in silico tools, including SpliceAI and Alamut Visual software [<a href="#ref-4">4</a>]. The use of multiple tools provides a more comprehensive assessment than any single tool, but it also increases the complexity of the analysis.

The practical approach is to use multiple tools and to interpret the results in the context of the variant location and the suspected mechanism. Strong concordance between tools increases confidence in the prediction, while discordance suggests uncertainty.

Observations and Measurements in Non-Coding Variant Interpretation

Measuring Splicing Impact

The measurement of splicing impact requires quantitative assessment of transcript isoforms. The RPGR study used RT-PCR to amplify variant and wildtype RNA and to investigate effects on splicing. The midigene assay confirmed that the variant led to the utilization of a cryptic acceptor site and a cryptic branchpoint motif, resulting in the excision of intron 10 and 90 bases of exon 11 [<a href="#ref-4">4</a>].

The measurement of splicing impact should include:

• Identification of aberrant transcript isoforms: Which isoforms are produced in the presence of the variant?
• Quantification of isoform ratios: What proportion of transcripts are aberrant?
• Assessment of functional consequences: Does the aberrant transcript lead to a frameshift, premature stop codon, or altered protein?

Measuring Regulatory Impact

The measurement of regulatory impact requires quantitative assessment of gene expression. Reporter assays measure the effect of a variant on transcription by comparing the expression of a reporter gene with the variant sequence versus the wildtype sequence.

The measurement of regulatory impact should include:

• Direction of effect: Does the variant increase or decrease expression?
• Magnitude of effect: What is the fold change in expression?
• Tissue specificity: Is the effect specific to certain tissues or cell types?
• Temporal specificity: Is the effect specific to certain developmental stages or conditions?

Recording Variant Classification Evidence

The documentation of variant classification evidence is essential for reproducibility and for future reclassification. The evidence should include:

• The variant coordinates and transcript context
• The population frequency data
• The in silico prediction results
• The functional assay results
• The evidence codes applied
• The final classification and the rationale

The nf-core documentation provides standards for reproducible workflow configuration and usage, which are relevant to the computational aspects of variant annotation and analysis [<a href="#ref-7">7</a>]. The Galaxy Training Network provides accessible workflow training and analysis tutorials that support reproducible analysis practices [<a href="#ref-8">8</a>].

Records and Measurements for Clinical Laboratories

Variant Classification Records

Clinical laboratories should maintain detailed records of variant classification decisions. The records should include:

• The variant identifier and genomic coordinates
• The gene and transcript context
• The evidence codes applied and the supporting data
• The classification and the date of classification
• The classification history, including any reclassifications

The ACMG/AMP framework requires that classifications be transparent and reproducible. The records should be sufficient for another laboratory to understand the basis for the classification and to reassess the classification if new evidence becomes available.

Quality Control Measures

Quality control measures for non-coding variant interpretation include:

• Confirmation of variants by orthologous methods: The carrier screening study confirmed pathogenic and likely pathogenic variants by orthologous methods before reporting them [<a href="#ref-3">3</a>].
• Validation of functional assays: Functional assays should include appropriate controls, including wildtype and known pathogenic or benign variants.
• Regular review of classifications: Variant classifications should be reviewed regularly as new evidence becomes available.

The Bioconductor project provides official package, workflow, installation, and reproducible genomic-analysis documentation that supports quality control in computational analysis [<a href="#ref-9">9</a>]. The Carpentries lessons provide foundational computing, data, shell, Git, and programming training that supports reproducible laboratory practices [<a href="#ref-10">10</a>].

Reclassification Procedures

Variant classifications should be reviewed when new evidence becomes available. The reclassification process should include:

• Monitoring of population databases for new frequency data
• Monitoring of disease-specific databases for new variant reports
• Review of newly published functional studies
• Reassessment of the evidence codes and classification

The RPGR variant was reclassified after functional analysis confirmed aberrant splicing. This reclassification demonstrates the importance of functional evidence in moving variants from uncertain significance to pathogenic [<a href="#ref-4">4</a>].

Common Failure Patterns in Non-Coding Variant Interpretation

Overreliance on In Silico Prediction

A common failure pattern is classifying variants based on in silico prediction alone. In silico prediction tools are useful for prioritization, but they are not sufficient for classification. The RPGR study used in silico prediction to identify the variant as a candidate for functional analysis, but the classification was based on the functional assay results [<a href="#ref-4">4</a>].

The consequence of overreliance on in silico prediction is misclassification. A variant that is predicted to affect splicing may have no effect in vivo, and a variant that is not predicted to affect splicing may have a significant effect.

Failure to Consider Tissue-Specific Effects

Non-coding variants often have tissue-specific effects. A promoter variant may affect expression in one tissue but not another. A splicing variant may affect splicing in one tissue but not another.

The failure to consider tissue-specific effects can lead to misclassification. A variant that has no effect in the tissue that was tested may have a significant effect in the disease-relevant tissue.

Incomplete Population Frequency Assessment

Population frequency is a critical filter for non-coding variant classification, but the assessment must be complete. The population databases must be checked for the specific variant, and the frequency must be interpreted in the context of the disorder prevalence and mode of inheritance.

The failure to assess population frequency adequately can lead to the classification of common benign variants as pathogenic.

Ignoring the Possibility of Cryptic Splice Site Activation

Deep intronic variants and exonic variants can create cryptic splice sites that are not predicted by standard splice site prediction tools. The RPGR variant created a cryptic branchpoint within an exon, which was not predicted by standard tools [<a href="#ref-4">4</a>].

The failure to consider cryptic splice site activation can lead to the misclassification of pathogenic variants as benign or of uncertain significance.

Limitations of Current Approaches

The Interpretation Burden

The interpretation burden of non-coding variants is substantial. The early-onset colorectal cancer study identified 1,042 intronic or non-coding variants and 44 3' UTR variants in 125 patients. The majority of these variants were classified as variants of uncertain significance [<a href="#ref-2">2</a>].

The interpretation burden is a practical limitation of current approaches. The resources required for functional analysis of every non-coding variant are not available in most laboratories.

The Lack of Functional Assay Standards

Functional assays for non-coding variants lack standardization. Different laboratories use different assays, different experimental conditions, and different interpretation criteria. This lack of standardization makes it difficult to compare results across laboratories and to apply evidence codes consistently.

The Limited Understanding of Regulatory Grammar

The regulatory grammar of the genome is not fully understood. The rules that govern transcription factor binding, enhancer activity, and splicing regulation are complex and context-dependent. This limited understanding constrains the development of accurate in silico prediction tools and the interpretation of functional assay results.

The Challenge of Reclassification

Variant reclassification is challenging because it requires ongoing monitoring of new evidence and reassessment of classifications. The resources required for reclassification are substantial, and many laboratories do not have the capacity for systematic reclassification programs.

Safety and Regulatory Context

Clinical Reporting Requirements

The reporting of non-coding variants in clinical settings is subject to regulatory requirements. The carrier screening study reported only pathogenic and likely pathogenic variants after confirmation by orthologous methods. This approach ensures that reported variants meet a high evidence standard [<a href="#ref-3">3</a>].

The reporting of variants of uncertain significance is more nuanced. Laboratories may report variants of uncertain significance with appropriate caveats, but the reporting must be clear about the uncertainty.

The Role of Professional Guidelines

Professional guidelines, including the ACMG/AMP guidelines, provide the framework for variant classification. The guidelines are updated as new evidence becomes available, and laboratories should stay current with guideline updates.

The ClinGen Sequence Variant Interpretation recommendations provide additional guidance for specific variant types and evidence codes. The RPGR study used ACMG/AMP and ClinGen Sequence Variant Interpretation recommendations for variant reclassification [<a href="#ref-4">4</a>].

The Importance of Multidisciplinary Evaluation

The interpretation of non-coding variants requires multidisciplinary expertise. The inherited retinal disease review emphasized the need for multidisciplinary evaluation and standardized outcome measures. The expertise required includes clinical genetics, molecular biology, bioinformatics, and disease-specific knowledge [<a href="#ref-1">1</a>].

The review also noted that molecular confirmation has become essential for access to novel gene-directed therapies, exemplified by voretigene neparvovec for biallelic RPE65 variants. This requirement underscores the clinical importance of accurate variant classification [<a href="#ref-1">1</a>].

Professional Escalation Criteria

When to Refer for Functional Analysis

Functional analysis should be considered when:

• A non-coding variant is identified in a gene with a strong prior probability of pathogenicity
• The variant segregates with disease in a pedigree
• In silico prediction tools support a functional impact
• The variant is in a region with known regulatory function
• The classification would change clinical management

The RPGR variant was referred for functional analysis because it was identified in a large pedigree with X-linked retinitis pigmentosa and in silico prediction tools supported a splicing impact [<a href="#ref-4">4</a>].

When to Seek Specialist Consultation

Specialist consultation should be sought when:

• The variant is in a gene with complex splicing patterns
• The variant is in a gene with multiple transcripts
• The functional assay results are ambiguous
• The classification has significant clinical implications
• There is disagreement between evidence sources

When to Report Variants of Uncertain Significance

Variants of uncertain significance should be reported when:

• The variant is in a gene with a known disease association
• The variant has some evidence of pathogenicity but does not meet the threshold for likely pathogenic
• The variant is in a gene that is relevant to the clinical indication
• The reporting would inform future testing or research

The reporting of variants of uncertain significance should include clear caveats about the uncertainty and recommendations for follow-up.

Frequently Asked Questions

What is the difference between a splice-site variant and a deep intronic variant?

A splice-site variant is located in the canonical splice consensus sequences at the intron-exon boundaries, typically the first two or last two bases of the intron. These variants directly disrupt the recognition of the splice site by the splicing machinery. A deep intronic variant is located deeper within the intron, away from the canonical splice sites. Deep intronic variants can affect splicing by creating cryptic splice sites, disrupting splicing regulatory elements, or altering branchpoint sequences. The RPGR variant described in this article was an exonic variant that created a cryptic branchpoint, demonstrating that splicing effects are not limited to canonical splice sites [<a href="#ref-4">4</a>].

How do I apply the PVS1 criterion to a non-coding variant?

The PVS1 criterion applies to null variants in genes where loss of function is a known mechanism of disease. For non-coding variants, PVS1 requires evidence that the variant causes loss of function. This evidence typically comes from functional assays that demonstrate aberrant splicing leading to a frameshift, premature stop codon, or complete loss of transcript. The RPGR study applied PVS1 after the midigene assay confirmed that the variant led to a frameshift and premature stop codon with no functional transcript predicted to remain [<a href="#ref-4">4</a>]. Without functional evidence, PVS1 should not be applied to non-coding variants.

What functional assays are appropriate for non-coding variant interpretation?

The choice of functional assay depends on the variant location and the suspected mechanism. Splicing assays, including minigene assays and RNA sequencing, are appropriate for variants that may affect splicing. Reporter assays are appropriate for promoter and enhancer variants. The RPGR study used midigene splice assays with gateway expression vectors [<a href="#ref-4">4</a>], and the carrier screening study used RNA-splicing assays for novel splice-site variants [<a href="#ref-3">3</a>]. The choice of assay should be guided by the specific hypothesis about the variant mechanism.

Can in silico prediction tools alone classify a non-coding variant?

In silico prediction tools alone are not sufficient for non-coding variant classification. The ACMG/AMP framework requires functional evidence for the strongest evidence codes, and in silico prediction is limited to the PP3 and BP4 codes. The RPGR study used in silico prediction tools to identify the variant as a candidate for functional analysis, but the classification was based on the functional assay results [<a href="#ref-4">4</a>]. In silico prediction should be used for variant prioritization and hypothesis generation, not for final classification.

How do population frequency data apply to non-coding variants?

Population frequency data apply to non-coding variants in the same way as to coding variants. The PM2 criterion requires that a variant be absent from controls in population databases, and the BS1 criterion requires that a variant have a frequency that is too high for the disorder. The carrier screening study filtered variants and classified them using the latest ACMG guidelines, with population frequency as a key filter [<a href="#ref-3">3</a>]. The interpretation of population frequency requires consideration of the disorder prevalence, the mode of inheritance, and the penetrance of the variant.

What is the role of disease-specific databases in non-coding variant interpretation?

Disease-specific databases provide information about variants that have been previously reported in association with specific disorders. The early-onset colorectal cancer study used the LOVD and ClinVar databases to classify variants [<a href="#ref-2">2</a>]. These databases can provide evidence for the PS4 criterion (prevalence of the variant in affected individuals) and can help identify variants that have been previously classified. However, the databases must be used with caution because the quality of the entries varies, and the absence of a variant from a database does not exclude pathogenicity.

How should I report a non-coding variant of uncertain significance?

A non-coding variant of uncertain significance should be reported with clear caveats about the uncertainty. The report should include the variant coordinates, the gene and transcript context, the evidence that was considered, and the rationale for the classification. The report should also include recommendations for follow-up, which may include functional analysis, segregation studies, or monitoring of new evidence. The carrier screening study reported only pathogenic and likely pathogenic variants after confirmation by orthologous methods [<a href="#ref-3">3</a>], but in diagnostic settings, variants of uncertain significance may be reported with appropriate caveats.

When should a non-coding variant be reclassified?

A non-coding variant should be reclassified when new evidence becomes available. The RPGR variant was reclassified from uncertain significance to pathogenic after functional analysis confirmed aberrant splicing [<a href="#ref-4">4</a>]. Reclassification should be considered when new population frequency data become available, when new functional studies are published, when the variant is reported in disease-specific databases, or when the ACMG/AMP guidelines are updated. Laboratories should have procedures for systematic review of variant classifications as new evidence emerges.

Related Bioinformatics Guides

• Detecting Structural Variants with Long-Read Sequencing: Methods and Considerations
• Foundation Models in Genetics: Opportunities and Challenges
• Metagenomic Assembly Overview: Challenges and Applications
• Deep Learning for Annotating Structural Variants in Viral Genomes
• How to Interpret Gene Set Enrichment Analysis Results

Related Clinical & Scientific Guides

• A Practical Guide to Detecting Antimicrobial Resistance Genes in Shotgun Metagenomic Data
• Computational Immunology: Modeling the Immune System
• How to Set Hard Filters for Germline Variant Calling: A Practical Guide to GATK Best Practices

References and Further Reading

This article is educational and does not replace validated analysis plans, institutional policy, clinical interpretation, or specialist review.